Show HN: Socratic – Automated Knowledge Synthesis for Vertical LLM Agents (github.com)

Socratic is an open-source tool that automates the conversion of messy, domain-specific documents (docs, code, logs) into compact, structured knowledge bases ready to feed vertical LLM agents. It orchestrates a 3-stage pipeline—ingest, synthesis, compose—where a user plus a terminal agent (codex) collaboratively extract candidate concepts, launch per-concept LLM agents to synthesize focused knowledge, and emit both human-readable text and JSON artifacts plus ready-to-use prompts for agent contexts. The repo includes a CLI workflow (create → ingest → synth → compose) and currently requires an OpenAI API key; gpt-5/gpt-5-mini are recommended models. This matters because building and maintaining high-quality, domain-specific knowledge has been manual, costly, and brittle. Socratic automates concept discovery and per-concept synthesis, producing modular, agent-ready knowledge that can accelerate domain agent development and reduce upkeep. Key technical implications: per-concept agentization enables parallel, targeted synthesis; structured JSON outputs make downstream indexing or retrieval-augmentation straightforward; and composed prompts let agents operate without ad-hoc engineering of context. Caveats include dependency on OpenAI models (cost and vendor lock-in), potential hallucinations requiring verification, and variable token costs for large corpora—so start small and validate synthesized knowledge before production use.